[Sluglug] new to the list

[cerise at armory.com]

On Mon, Oct 03, 2005 at 03:49:19PM -0700, Ignacio Solis wrote:
> What are you talking about? You suddenly jumped to talk about disk latency and
> about sectors.  When data is defragmented it is done so by grouping the data,
> but not into sectors (which btw currently might have no correlation to the
> placement on disk).  Data in contiguous space (addresses) on the disk are
> considered to be close to each other, which generally means close by wrt
> tracks/cylinders.

In response to my suggestion that the filesystem spread files out around the
disk, you responded:
> That doesn't make sense. If you want fast access (seek time) you want all the
> files close to you (the head), hence you move them close together.  You'll
> achieve nothing if you spread them out... other than increasing the seek time
> from one file to another.

Fast access tends to imply minimal disk latency.  Minimal disk latency in
this situation implies that you follow my recommendation in my last post.

You are correct that it is defragmented by grouping the data and attempting
to make the maximum use of clusters, however that only ensures fast access
of the file if you're in the right place at the right time which 
happens very infrequently.  In fact, it happens even more infrequently (as
I showed in my last post) if you group all frequently accessed files into
one location on disk as you indicated to be a poor idea with the last
sentence of that quote.
 
> Schedulers are a different beast, but they still benefit from files being
> close by.

False.  Schedulers attempt to optimize reads and writes based on the current
position of the head over the disk.  If the frequently accessed files are 
close by (read: localized to one region of the disk), then it will have a lot
of work for the disk when it happens to be in that region and almost no
work for it the rest of the time.  It runs the rather believable risk that
it will have too much work to do in that region and not complete all reads/
writes while the head is over that region.

However, if frequently accessed files are continuous and spread throughout
the surface of the disk, that work is spread over the course of the entire
rotation of the platter and is much more likely to be a workload which the
head can perform.

One might note that if only one file is being accessed by one application,
then this recommendation is no worse than yours.  In any situation where
multiple applications are accessing any number of files, then it is trivially
better than your recommendation as shown.

-Phil/CERisE